recent, Science, Science / Tech

Science and Data: Notes on a Misconception

Science would be impossible without evidence. But champions of science frequently portray data or evidence as the fundamental building blocks of scientific knowledge. They will say that scientific theories should be “evidence-based”—confirmed by data or supported by evidence. They will say that a good theory follows inexorably from the evidence, whereas a bad theory has little or no evidence to back it up.

But this familiar way of thinking about science is the misconception that the contents of scientific theories somehow emerge out of the data. In fact, as the physicist David Deutsch showed in his 2011 book The Beginning of Infinity, the contents of scientific theories are explanations of the data. And people, not data, are the source of those explanations.

The misconception that scientific theories are “based on evidence” has an important practical consequence: it steers science toward generating data instead of seeking explanations for the world. To aim science explicitly toward seeking explanations, we must clarify and overturn this misconception, and present an alternative vision of an exclusively explanatory scientific ethos.

Creativity and Explanatory Science

The theory that science is based on evidence derives from a philosophy known as empiricism. Empiricism is the idea that we obtain knowledge through our senses, by gathering data. This theory of knowledge, or epistemology, is associated with some of history’s most prestigious scientists and philosophers, from the Enlightenment philosophers John Locke and David Hume to the logical empiricists and quantum theory pioneers of the twentieth century. Today, empiricism endures in the prevalent idea that evidence is the foundation of science.

But despite its enduring influence, empiricism belongs to an obsolete worldview. By claiming that knowledge derives from the information streaming into our senses, it postulates the existence of direct, or manifest experience. Yet, all of our experiences are mediated through layers of complex physical processes in our nervous systems. And we understand those processes not through perception, but indirectly: through our explanations.

Scientific theories do not emerge from data. They emerge from creative thinking, which we can apply to generate explanations of that information streaming into our senses. Creativity makes it possible to ask a question, guess an answer, criticize the guess, and then ask another question. Science consists of using this ability to explain the evidence of the world. These explanations are not derived from the data; they are conjectured to account for it.

Ironically, the idea that data provide a basis for scientific theories fails to account for the contents of those very theories. Scientific explanations typically describe entities and processes that our senses never detect. For example, nobody has ever observed Earth’s slowly convecting mantle layer, yet we assert its existence to explain various phenomena that we do see, such as earthquakes and volcanoes.

Even if we could see Earth’s mantle outside of our windows, we would still need to explain what we saw. And this would require creativity as well, because data do not explain themselves to us. As Deutsch remarked during a 2009 TED Talk, creativity gives us the ability to “explain the seen in terms of the unseen.” We see volcanic activity. We explain it by describing a hot, dense, molten world beneath our feet.

Creativity lies behind every scientific discovery: past, present, and future. But empiricism is not the only culprit idea behind the misconception that scientific knowledge is based on evidence. Empiricism is closely related to an even deeper and more persistent misconception, which is not just about science, but all knowledge: it is called justificationism.

Unjustified Knowledge

“Knowledge” has various meanings, but in philosophy by far the most common and influential concept is of knowledge as a justified true belief. The difficulties inherent in this tripartite concept have absorbed justificationist philosophers for centuries. In justificationism, the essential question is: What constitutes a justification for a belief? And justificationism’s central weakness is that no matter how you answer, you face an infinite regress that repeatedly demands: What justifies your answer?

The justified-true-belief idea has impeded progress in epistemology by making it about justification rather than knowledge. As a result, many philosophers have ignored more interesting and substantive epistemological problems, such as how knowledge is physically possible, and what that could tell us about its limits and potential reach. Neglect of such problems is not surprising, because although they are about knowledge, they are not about justified true beliefs.

Rejecting justificationism, Karl Popper introduced the more fruitful idea of objective knowledge in a 1972 book by that very title. Objective knowledge is a certain type of information: information that is in any way true or useful. Initially, this concept seems vague or simplistic. Yet many of our best explanations of the world automatically entail this concept of knowledge.

For example, the theory of evolution states that an organism’s genes embody true and useful information about how to self-replicate in a specified environment; and similarly, airplanes fly because their designs contain true and useful information about how to achieve lift. The true and useful information that is encoded in gene sequences and airplane designs are both prime examples of objective knowledge.

Crucially, whereas the justified-true-belief concept entails that knowledge requires a knowing subject—because it requires belief, a subjective state of mind—objective knowledge can exist without a knowing subject. It refers only to the knowledge: the impersonal information content. Once a piece of scientific knowledge is created, its contents exist objectively, instantiated in some physical object (e.g., a brain, book, or computer), independent of anyone’s beliefs.

Empiricism holds that scientific knowledge is based on data or evidence; justificationism claims that all knowledge consists of justified true beliefs. Together, these ideas lead to anthropocentric and reductionist portrayals of science, in which we rely on sense data to justify our beliefs about the world. But scientific knowledge is not a class of evidence-based true beliefs. It is better understood as a collection of conjectural explanations, created by people attempting to understand the world.

The Relationship between Evidence and Explanations

Creating objective knowledge by forming scientific explanations does not involve empirical justification. In fact, it requires no justification at all. However, creating explanations does require imagination, because our explanations of the world are filled with entities and processes we have never observed. People have created this explanatory knowledge by the process previously described: using creativity to ask a question, guess an answer, criticize the guess, and then ask another question—a process that is fueled by imagination, not by evidence.

But if evidence cannot yield or justify scientific theories, then what role does it play in the growth of scientific knowledge? Ironically, the empiricist focus on evidence has kept evidence from fulfilling its basic function in science: to criticize and falsify theories.

And criticism is essential. Without it, explanations cannot improve, and science cannot progress. But how do we know that the function of evidence in science is to criticize theories? For one important reason: Although theories cannot be logically derived from evidence, they can be logically contradicted by the results of experiment or observation.

Thus, the idea that scientific theories are “evidence-based” misconstrues the logical relationship between evidence and explanations. Conversely, viewing science as exclusively explanatory illuminates why evidence is so crucial to scientific progress.

If scientific knowledge consists of imaginative and conjectural explanatory theories about the world, then it is invalid to challenge a theory by demanding evidence for its propositions, because all of its propositions are creative guesses that originate in the mind, not in the evidence.

However, providing evidence that contradicts a theory’s proposed explanations constitutes a severe and genuine challenge to the theory. Such a challenge might reveal flaws in the theory and thus help us to improve it. To be sure, the critical discussion of any scientific theory does not revolve purely around evidence, and rational criticism in science need not involve data. But the possibility of adducing empirical evidence to criticize a scientific explanation is what distinguishes science from all other fields of human explanation.

Science would be impossible without evidence. However, to improve our scientific knowledge, we must conjecture explanations of the evidence, not attempt to derive theories from it. By applying creativity and seeking conflicts between evidence and explanations, we can continuously improve those explanations. We should continue to prize evidence, because although it cannot provide a basis for scientific theories, it has always provided something essential to scientific progress: diverse, mysterious, and beautiful explicanda.

 

Lucas Smalldon is a science writer in the Washington, DC area. You can follow him on Twitter @reason_wit_me

Feature image: Galileo showing to John Milton the markings on the moon.

Comments

  1. I feel like there was a good point lurking somewhere in there, but I didn’t see it made.

    Perhaps what the author was dancing around is that when one starts with a correlation, then postulates a (ideologically-flattering) cause and declares victory, he has neglected to consider evidence that he is wrong. Whereas, if one starts with the theory, pre-evidence, he might be more likely to formulate experiments that could debunk himself?

    Of course, quite a lot of “science” these days is ideology trying to rationalize itself.

    To me, the issue is falsifiability. Falsifiability is the bedrock of scientific process - it must be possible for a theory to have failed a test for the passing of said test to mean something. Much modern political discourse is not falsifiable, including most all privilege/oppression narratives and accusations of bigotry. This is a fundamental flaw.

    It would be nice if we had a concise term for the “evidence” that exists in favor of a notion that is clearly debunked by other evidence. For example, one might observe that a person who was vaccinated has autism and conclude that vaccines cause autism, even as mountains of evidence to the contrary exist. I would like to be able to refer to that initial piece of “evidence” (the sample size of one) as an X, where X is known to mean “evidence in favor of a conclusion that is nonetheless wrong.”

  2. Where’s the evidence that such conflations of science and evidence/data are a problem/concern that needs to be addressed? :astonished:

    Scientific theories without evidence are not valid theories, just hypotheses. A theory requires a well established system that is based on evidence, and that evidence is in collected data.

    Great hypotheses lead to the collection of data that tend to support or refute it. Sometimes, the data leads to new hypotheses, so science can advance by analyzing existing data as well as scientific hypotheses cause people to seek ways to gather data.

    It’s unclear to me what the real underlying issue or idea is in this piece. Science does require imagination, but without data/evidence, it remains an unproven hypothesis, even when there’s some evidence (typically in mathematics, based on calculation on data) like the multiverse life on other planets, dark energy/matter, etc.

  3. Thanks for this great explanation of how science should be done - and the dangers of doing it contrary to Popperian principles. More particularly, I think it is a mistake to think that science proves anything at all.

    In the Popperian ideal, as I understand it, science is the business of generating tentative theories (hypotheses) to explain observations of the natural world and the results of contrived experiments. (Below, “observations” includes both.) The theories are debated and evaluated against the observations. Part of this is debating how the actual events inside and outside experiments are interpreted mechanically and intellectually to form what we regard as observations. This is where it gets tricky - hence this longer than average comment.

    The result, at any one time, is that there are 1 or more theories, where each is viewed, by each participant, as more or less complete an explanation, or probably or certainly false. Such assessments are subject to their own levels of doubt by each individual, and different individuals have different opinions on the theories.

    In mathematics (this is my conception, not anything I read), one can prove and disprove postulates, the validity of equations etc. because mathematics consists of an entirely human-generated, formalised, set of axioms, principles etc. which people use as the criteria for assessing the validity of any new mathematical theory. There might be some debate about which axioms to apply, and how, but the process is entirely mental and rule-based (albeit aided by computers to deal with excessive complexity) within the minds of the participants, with words, diagrams, formulae etc. flowing between them. Mathematics does not involve the physical world: observations or experiments. I think mathematics is far overrated in physics.

    Strictly speaking, Popperian science never proves anything. It amasses observations and experimental results, together with their low-level interpretations - which should always be subject to potential revision. It generates and debates theories. In the ideal outcome, one theory is widely regarded as explaining everything which seems to need explanation, with any other theories having been found to be inadequate, due to them failing to explain one or more observations.

    Not all theories are scientific. Only those which explain observations and have predictive power are scientific. A scientific theory is one which predicts the results of observations and experimental results in the future - and a successful scientific theory is one which is judged to have done this very well. It is more impressive for the theory to have originated at time X, while at time Y well beyond that, we agree that the observations made between X and Y are found to be entirely in accordance with the theory. The more scientific the theory, the more particular its predictions.

    Scientific theories are also evaluated according to their elegance - which is a subjective matter and so highly subject to debate and current beliefs. Elegance involves reducing the number of elements in the theory to the minimum required to do the job. I think elegance also involves the internal elements of the theory corresponding to processes which we think actually occur in the physical world. This, inevitably, involves some degree of testing theories against other theories.

    If those theories used as the basis for evaluating elegance are in fact (and strictly speaking no-one really knows absolutely whether they are) accurate, then this is a good way of evaluating a theory’s elegance. It amounts to evaluating how coherent the theory is with all other well accepted theories, which current opinion holds are the best available.

    Ideally (and I assume Popper wrote this) scientific theories are evaluated solely by their ability to explain observations.

    However, this is where it gets messy. Some widely believed theories are in fact incorrect. Maybe no-one knows this at time X, but at time Y, later, most people will agree these theories have been falsified due to the inability to predict observations. These observations might only have occurred after X, but in many cases the observations predate X.

    So a widely believed scientific theory can be entirely wrong. Maybe one or a few people know about this, and are certain about it, and will be generally accepted by most people at time Y to have been entirely correct about this all along.

    One classic example of this is Copernican Revolution - heliocentric theory replacing the earlier geocentric theories developed by Ptolemy. (Another is the 1960s acceptance of plate tectonics - see “Drifting Continents and Shifting Theories” by H. E. Le Grand.) As more detailed observations of planetary motion accumulated, the Ptolemaic system was elaborated with more and more epicycles until the result became ridiculously complex, and so inelegant. When the main body of scientific opinion shifted in favour of the Copernican theory, this was what is now known (thanks to Thomas Kuhn) as a paradigm shift.

    So a paradigm is one or (more likely) multiple interlocking scientific theories. At time X, most people consider it to be truthful - a good, if not perfect, explanation of Nature. At time Y, everyone - or almost everyone - agrees that the theory was false. Usually, this only happens when, by time Y, another theory has been developed and found (in the opinion of scientists of the day) to be superior in every important respect to the older, now discredited theory.

    Ideally, scientists would be happy to reject a falsified theory in the absence of any so-far unfalsified and so more attractive theory. However, I think this rarely occurs.

    Some areas of science are highly amenable to experiment. So the theory that gastric ulcers are caused by excessive acid was easily displaced once experiments showed that the cause was bacterial infection - Helicobacter pylori.

    Others are less so. Cosmology is not amenable to experimental investigation. Neither is human evolution. Evolution in general is slightly amenable to experiment, or at least observations over decades, such as with moths changing genetically so their colour matches those of trees darkened by industrial soot.

    Some areas of science could be investigated experimentally, but ethical or practical concerns mean the experiments are never conducted. If ethics were of no concern, one could experimentally deprive pregnant women and/or their newborns from nutrients and care, or subject them to potential poisons and ill-treatment - and so generate robust observations which would help in evaluating numerous theories which would be practically valuable.

    The science of atoms, sub-atomic particles and electromagnetic radiation is highly subject to experimental investigation, but there can be problems with interpreting the data which arise from these, due to such interpretation being based on potentially faulty paradigms.

    In cosmology, because almost all participants believe in the Big Bang Theory as fact (BBT, not just theory, but such a strongly established theory that they cannot imagine any other theory ever replacing it), this has profound effects on how they interpret their observations.

    For instance, the one observational 2DF program collected spectra and thereby calculated redshift for thousands of galaxies and quasars. Because the quasar redshifts are higher, they were assumed (according to BBT) to be far more distant than all the galaxies, so the observations were subject to different analysis pipelines and were published as separate databases, with separate journal articles.

    Ultimately, assuming science progresses well, any such mistakes will be fully exposed and widely accepted as being part of a by then discredited pattern of thinking and analysis according to a false paradigm.

    A central BBT tenet is that the redshift (lower frequency, longer waves) of light (and more generally electromagnetic radiation = EMR) from distant (beyond our galaxy) objects is entirely due to Doppler shift - their motion away from us. In the BBT, the relative motion is proportional to distance, so redshift combined with the currently agreed Hubble constant (~68km per second per megaparsec, though this is subject to debate) results in a distance for the observed object.

    This seems to be, approximately at least, true for observations of galaxies, since there is the expected anti-correlation between their redshift-BBT-derived distances and their observed brightness (amount of light arriving in our telescopes).

    This correlation does not apply to objects identified as quasars. Quasars have a flat spectrum (completely different from that of galaxies, which have star-like spectra with a black-body-radiation-like central peak in the visible) best explained by synchrotron radiation of matter (plasma) spinning around and into the accretion disc which surrounds a black hole.

    This is properly regarded by a few people as a disproof of the BBT - however it is generally swept under the carpet. To admit this problem would be to admit that the quasars are not at their redshift-BBT-derived distances. To admit this would be to admit that the redshift EMR from distant objects is partially, or perhaps wholly, caused by as-yet unknown processes (tired light) which occur as the EMR traverses the sparse (such as one electron and proton per cubic metre) plasma of intergalactic and inter-galaxy-cluster space.

    Generations of cosmologists and astronomers have earned their living from taxpayer and philanthropist funding, to a large degree based on the project of understanding the Big Bang. It would take a monumental paradigm shift to admit this was a mistake - or even that the theory had serious and potentially fatal weaknesses.

    Photons - light quanta - were proposed by Albert Einstein in 1905 and his theory was widely disbelieved until the mid-1920s, when some perplexing observations of X-ray diffraction from crystals could most easily (but not necessarily correctly) explained by the X-rays themselves being particles, rather than waves. Gamma rays, X-rays, ultraviolet, visible, infrared, terahertz/millimetre wave, microwave, VHF, short-wave, medium wave and long-wave (500khz and below) are all electromagnetic waves. So unless someone can show that at the gamma and X-ray high frequencies, there are particle properties, but not at optical and radio frequencies, then it seems (to most people) reasonable to assume that all EMR is particles - photons. Yet even Einstein, who won the Nobel prize for this article, admitted a few years before his death that he had no idea what light quanta were.

    The acceptance of photons was a paradigm shift, replacing classical waves as the explanation for EMR - or rather accepting that scientific criteria for theories should be loosened on the basis that EMR could possess both particle and wavelike properties, though supposedly not at the same time.

    A well recognised unsolved problem in astrophysics is the heating of the solar corona - to 1 million Kelvin, while the photosphere is about 5500K. Most scientists believe entirely in the BBT absolutely, despite being unable to explain a profound mystery on our doorstep.

    I believe this is entirely due to their inability to think of light as classical waves - they consider it only as photons. Photon theory is that EMR cannot interact with the corona, so the idea that sunlight heats the solar corona is considered for a moment, and then rejected.

    This is a grave mistake, since it is evaluating - and absolutely rejecting from further consideration - one or more scientific theories entirely on the basis of theory. This is at odds with Popperian principles. This highlights the difficulty we have distinguishing between the actual processes of nature and how we interpret them.

    The idea that sunlight heats the solar corona is not falsified by observations. Once the Sun’s atmosphere gets above about 1800km from the photosphere, drops in density and rises in temperature to 1MK or so, the heating and acceleration of the corona does correlate with the intensity of the sunlight - initially 64 megawatts per square metre. (The corona is as sparse as a high vacuum on Earth, so it doesn’t take much energy per volume to heat it.)

    Thinking of light as waves, it is easy to think of the heating mechanism in a suitably hot, sparse plasma. Electrons move faster than protons, since they are much lighter. Their instantaneous velocity is added to and subtracted from by the sunlight’s electrostatic fields. As long as the electron continues moving in a straight line, these increments and decrements balance exactly and the average velocity is unaffected. Protons are hardly affected by these fields since they are 1836 times heavier.

    When the electron collides with a proton, its trajectory deviates, so it is no longer subject to the same exactly (on average) balancing pattern of positive and negative increments in its velocity which it gets from and gives back to, the electrostatic fields of the sunlight. It is suddenly inserted into a new trajectory with a different, also average zero, fields which are partially or wholly uncorrelated with those of its original trajectory. Therefore, the electron brings some amount of that velocity increment or decrement it had from the initial trajectory to the collision, depending on the angle by which its new trajectory deviates from the that of the original. The kinetic energy they bring to the collision is proportional to the square of their incident velocity with respect to the relatively stationary proton.

    If the velocity is V and the increment I, with a positive increment the kinetic energy is proportional to (V + I)^2 and the negative increment is proportional to (V - I)^2. While the velocities average exactly to the original unperturbed velocity, the kinetic energies average so something slightly higher than the unperturbed kinetic energy. (The average of 7 squared and 5 squared is 37 - higher than 6 squared).

    So the average kinetic energy of electrons at the point of collision with protons is raised by the sunlight. That kinetic energy is shared by the subsequent trajectories of the electron and proton - and it is never returned to the sunlight. Sunlight heats the solar corona.

    Ideally I will extend this theory of coronal heating by sunlight to predict exactly how much would occur according to the brightness and other characteristics of the light, and the temperature, density etc. of the plasma. This would make it more strongly predictive and so more amenable to being found faulty against our best estimates of the actual heating of the corona at different distances.

    Similarly, I expect the barrier to devising a tired light theory of redshift in the intergalactic medium is that most people are thinking in terms of photons - which according to theory don’t interact with it at all. I believe photons do not exist. (I have tentative non-photon theories for this redshift, the photoelectric effect and the black body spectrum - but not yet for X-ray crystal diffraction.)

    Popperian science is the ideal. Kuhn examines what actually happens, with human failings of being swept up in prevailing, sometimes false, paradigms. These paradigms affect our interpretation of observations, such that what we think are the observations may actually be incorrect interpretation of natural and experimental phenomena. People cannot work in a field, pay their mortgages, care for their loved ones etc. if they don’t accept the currently prevailing paradigm.

    Perhaps the most infuriating thing about (ultimately judged, at time Y, to be) false paradigms is that no-one, or almost no-one, can tell they are false at time X. Everyone assumes they are true. At time X no-one thinks of what might later be known as, for instance, the “photon paradigm”.

    Rejecting new theories directly due to their conflict with currently accepted theories is plain bad science. Rejecting them due to our current understanding of observations amounts to the same thing when our understanding is incorrect due to false prevailing theories. This is much harder to guard against. So we need to keep our imaginations wide open.

    It is all downhill from there - and here we see the false (I argue) photon paradigm (the Einsteinian foundation of quantum mechanics and so all contemporary physics) locking in the misunderstanding of EMR and so making it impossible for most people to think their way out of the Big Bang Theory. This is despite numerous contradictions between theoretical predictions and observations for both theories.

    We need some class of theories to regard as fact, for our own sanity. The Earth orbiting the Sun is one of these. I believe it will never be disproved, so I regard it as a fact.

    Science should be thought of along Popperian lines for its ideal embodiment, and according to Kuhn and others regarding its human failings. It is a complete mistake to think that science, at least in the Popperian view, proves anything.

    Yet we need to assume some things are facts so we can make decisions - and science, while it proves nothing, is the best method we know of creating theories which are likely to be reliable explanations of Nature.

    Science has nothing to do with human value, morality etc. - other than, ideally, providing a good explanation for the Natural world which we can use as a basis for these.

    Also, until and unless mainstream science admits the existence of dimensions, forces, stuff or whatever beyond the currently accepted 3 dimensions, time, matter, gravity, electomagnetism and de Broglie’s wavelike properties of momentum and matter (far more perplexing than all the rest) etc., it won’t be able to develop hypotheses regarding the more tenuously accessible aspects of the Universe many or most people believe in - precognition, ESP, psychokinesis, things that go bump in the night, souls and the afterlife.

  4. @Dok

    I’d say that listening to or reading any mass -media report that cites “science” as an authority will provide ample evidence of the problem. Climate change hysteria is the obvious example. It’s presented as an exceedingly simple theory: “increased carbon dioxide in the atmosphere leads to exponentially increasing catastrophe.” The author’s post above easily shows how this isn’t very strong scientifically, since it’s easy to observe all kinds of data that contradict the theory. There are stronger scientific variations of the theory such as “increasing carbon dioxide in the atmosphere leads to less infrared radiation from Earth to beyond Earth, all other variables held constant.” That version can at least be tested against observations, and also has a lot of creative thinking behind explaining it. But it has no political value. Maybe a less controversial example would be in “big data” and medicine. By data mining existing medical data all kinds of correlations will be found. The best scientists will combine the data with deep understanding of amazingly complex cellular, molecular, and organismal biology and develop a theory such as: “[promising treatment] would cure nearly 100% of [blank disease] in humans with these three SNPs: ____, ,. We abandoned the treatment previously because we didn’t know that only 5% of humans have those Gene combinations. That makes sense because of [this complex cellular process]. Based on this new theory we should try it again on that limited population.” That latter part is what the author is talking about. Theories like that exist, but they are scarce because the vast majority of “science” communication is really just noise generated from news about data.

  5. In other words science is often provisionary and can be best described as the use of the scientific method to evaluate theories.
    When I hear the word ‘‘evidence’’ bandied about in these discussions of science I often wonder whether anyone using the term has any legal qualifications. The rules of evidence worked out over hundreds of years are a very good guide to finding the truth. I often think that they should be sed to evaluate the claims of the global warming believers.

  6. I agree with DOK and GeorgeQTyrebyter.

    It feels like the author aims to devalue the term “empiricism” in favour of “creativity” and “imagination” - and debunks his own arguments on the way.

    “…to improve our scientific knowledge, we must conjecture explanations of the evidence, not attempt to derive theories from it.”

    Aren’t the former and the latter basically the same?

  7. The article seems to belabour the obvious which is that the key role of data in science is to falsify theories but he oversteps the mark in two ways. First by saying that evidence does not lead to theories. This is wrong evidence or data alone does not inexorably and directly lead to a theory, human imagination and creativity is needed but the input and inspiration to that is the data - observations about the world. Second he seems to think that this supposed misapprehension that theories spring direct from data is a serious problem and impediment to science. It isn’t I don’t think it is a common view amongst scientists at all. I have nevre rmet anyone who thinks such a thing, which is evidence to falsify this theory, although admittedly not very good quality evidence.

    There is a big problem with epistemology in the modern world but it is not this relatively suble issue but the post-modern/gender studies/social sciences view that knowledge and science are simply social constructs that reflect power relationships. This is a big problem and has deeply corroded large areas of academia and has started to cause huge problems in society.

    In certain areas paticualrily around child protection, sex crimes, domestic and other violence policies are normally advocated based on theories which runs directly counter to the evidence. Any objections are met with claims that arguments based on logic, statistics and evidence are constructs of the oppressive power structures within society.

    The rather subtle error of epistomology highlighted by the author pale into insignificance compared on a direct assault on the need for rationality, consistency and the abiliity for falsification. I would rather overstate the power of data than deny its utility at all.

  8. The article wasn’t very good, because it seems to be based on a number of misconceptions, ironic for the title. Also, it leaves the reader with a big “so what?” feeling at the end.

    “Ironically, the idea that data provide a basis for scientific theories fails to account for the contents of those very theories. Scientific explanations typically describe entities and processes that our senses never detect.”

    The author is misusing the word data to mean sensory stimulation, against the more usual meaning “observations and measurements about the world around us”. No, people can’t see the convective mantle of the earth, but we do have precise data on the time it takes for various waves to travel through the globe, their relative strengths, and more importantly we know that transverse waves propagate differently to compression waves, which is one the things that differentiates solids from fluids. The real theory at heart here is that of wave propagation in solid and fluid mechanics, which is established based on a combination of physics and mathematics. It is based on pre-existing physical theories (Newtonian mechanics) and laboratory experiments (i.e. clean data).

    I think a better conception of the whole thing is more like this: science has 3 main parts. One is the task of finding good data, i.e. experimentation and observation. The second is modelling, which is figuring out how theory can apply to various situations, and what is a good description of a situation. This is usually called theory but I prefer modelling as a term because a model is a tool to understand, without necessarily being assumed to have some deeper sense of truth to it. Newtonian physics is a good example - it is not “true” in any strict sense, but it is good enough a model for everyday usage in most applications.

    The third part is computation and simulation. This is the task of putting a model to predictive use. In reality, the vast majority of scientific models are not easy to turn into predictions without serious computer simulations, unless one assumes very simple scenarios. This third part is probably the fastest growing in modern science. Typically, these computations use approximations to the real model, and studying the accuracy or fidelity to the full model is a major part of modern science. Typically this involves substantial and deep mathematical theory of the behaviour of the underlying model and of computer algorithms.

    Overall, modern science has its basis in all of these, i.e. it wouldn’t be possible to do science without all three. Different fields have different emphases, e.g. medicine is almost entirely experimentally driven, whereas engineering is largely driven by modelling and computations. But to try to elevate one above the other seems me somewhat irrelevant and silly.

  9. As I understand the article, theories should first be proposed and then tested. The data collected during the testing will either help confirm the hypothesis or falsify it. (So far so good). This causes the author to claim that the data must always follow the theory (Problem). Sometimes data is simply collected temperature readings, rainfall amounts, actuary tables, ect… In the process of collecting data the researcher may begin to notice certain patterns and thus propose a theory. Therefore I propose that the notion the data should always follow the theory is falsifiable.

  10. Here’s a good example of the new “science”:

    https://www.salon.com/2019/10/10/thanks-to-trump-std-rates-hit-a-record-high

    The author ought to be penning an “I made it all up” apology similar to the one that recently appeared in Quillette, but I doubt that will happen. There is no attempt at evidence linking the STD increase to Trump, and there’s certainly no attempt at falsification. The author simply saw news (STD rates are up) and decided to manufacture a way that it was Trump’s fault. The results were fed to the tribe, which ate it up enthusiastically.

  11. Not only is that a stupid quibble, but it is false.

    From the AMS website:

    Math is not science. Sciences seek to understand some aspect of phenomena, and is based on empirical observations, while math seeks to use logic to understand and often prove relationships between quantities and objects which may relate to no real phenomena. Scientific theories may be supported by evidence, but not proven, while we can actually prove things in math. On the other hand, math is like science, and emphasizing the difference may really work against math.

    https://blogs.ams.org/phdplus/2017/04/17/math-is-like-science-only-proof-y/

  12. Regrettably, all you are doing is exposing your ignorance. Math is the language of science. It is not, however, science, and works in entirely different ways. In math, you prove things. In science, nothing is ever proven. Theories are considered “accepted”, but never proven. This is why, after 150 years, evolution remains a theory. It is not proven.

    Nothing in science involves proof. No paper, written by a responsible scientist, claims to prove anything. All papers examine evidence, which either supports or fails to support the hypothesis.

    Mathematics is a language, about abstract concepts which are unconnected to reality. Science involves reality.

    You should learn something about science. I regret to observe that you seem unclear on fundamental issues.

  13. Another discussion of the difference between science and math:

    Misconceptions about the nature and practice of science abound, and are sometimes even held by otherwise respectable practicing scientists themselves. I have dispelled some of them (misconceptions, not scientists) in earlier posts (for example, that beauty is in the eye of the beholder, beauty is only skin-deep, and you can’t judge a book by its cover). Unfortunately, there are many other misconceptions about science. One of the most common misconceptions concerns the so-called “scientific proofs.” Contrary to popular belief, there is no such thing as a scientific proof.

    Proofs exist only in mathematics and logic, not in science. Mathematics and logic are both closed, self-contained systems of propositions, whereas science is empirical and deals with nature as it exists. The primary criterion and standard of evaluation of scientific theory is evidence, not proof. All else equal (such as internal logical consistency and parsimony), scientists prefer theories for which there is more and better evidence to theories for which there is less and worse evidence. Proofs are not the currency of science.

    Proofs have two features that do not exist in science: They are final , and they are binary . Once a theorem is proven, it will forever be true and there will be nothing in the future that will threaten its status as a proven theorem (unless a flaw is discovered in the proof). Apart from a discovery of an error, a proven theorem will forever and always be a proven theorem.

    In contrast, all scientific knowledge is tentative and provisional , and nothing is final. There is no such thing as final proven knowledge in science. The currently accepted theory of a phenomenon is simply the best explanation for it among all available alternatives . Its status as the accepted theory is contingent on what other theories are available and might suddenly change tomorrow if there appears a better theory or new evidence that might challenge the accepted theory. No knowledge or theory (which embodies scientific knowledge) is final. That, by the way, is why science is so much fun.

    Note also the distinction in this discussion between math/logic and science.

  14. I’m not sure I understand this article, but I want to make two broad points, one a quibble about the geology example, the second about the philosophical point.

    The existence of earthquakes and volcanoes are not evidence for a convecting mantle. As someone above alludes to, the evidence for that is primarily from geophysics (although the convecting mantle is solid, not fluid. It just exhibits ductile deformation, as opposed to the brittle deformation of the shallower layer). The convecting asthenosphere (the layer that deforms ductily) is connected to some volcanism through the mid-ocean ridge and subduction system, which are the surface manifestations of convection, bringing material up and down, respectively. However, we also have ocean island volcanism, like at Hawaii, which is not really related to plate tectonics or mantle convection, so volcanos themselves are not evidence for convection. Similarly, earthquakes are related to plate tectonics but not really evidence for the convection going on much deeper.

    As to the philosophy, I think there’s some confusing concepts and invitations for logically fallacies. First there’s this distinction drawn without a real difference:

    Either way, you start with the data and seek to explain it. That we are limited by our human faculties and utilise creativity in generating explanations is a given.

    This is an astonishing statement. I think I would be fired from my PhD if I said this to my supervisors. Of what use is your “explanation” if it is untethered to empirical justification? How would you ever hope to get this published? This just seems like an invitation to engage in ad hoc reasoning and non-sequiturs. Every statement must be supported by the data or it isn’t science and no one will believe you.

    Maybe this approach is acceptable in post-modernism infected humanities departments, but it is totally unacceptable in real science.

    Of course theories can be derived from evidence. You analyse the oxygen isotope ratios of a mantle rock, you get values outside the range that can be explained by high-temperature fractionation, and you theorize a low-temperature origin or influence. Why should reasoning like this not be logical?

    I can only imagine the amount of nonsense that would be accepted as scientific knowledge if we removed the burden of substantiating our theories. If the only purpose of evidence is to falsify hypotheses, that gives express permission to advance unfalsifiable claims.

    Creativity is undoubtedly important to science, and society would be better off if it was understood just how much personal biases have room to play when you have a huge amount of data and many ways of treating it, but it is a step too far to say scientific knowledge doesn’t need to be justified by data

  15. Yes, sorry, I would have tagged you if I had remembered it was you or been motivated enough to scroll back. I appreciated that you brought it up to start with. It’s a very common misconception that the mantle is fluid, I had to bring it up because this is the closest to my field of expertise we’ve ever gotten to on Quillette. Got to savour actually being qualified to say something, instead of my usual spouting off :joy:

Continue the discussion in Quillette Circle

80 more replies

Participants

Comments have moved to our forum

96 Comments

  1. Pingback: Feuerbringer rettet die Demokratie « Feuerbringer-Magazin

Comments are closed.